Applications of Control Theory in Biomedical Engineering

Cheng-Ning Huang, Hung-Yuan Chung*

Department of Electrical Engineering, National Central University,
Chungli, Taiwan 320, R.O.C.
Tel:(03)4227151 ext.4475
E-mail: hychung@ee.ncu.edu.tw
Abstract
In this paper, we present an overview of various
applications control theory in biomedical engineering
application. The goal of this paper is concerned with
the problem of control theory in biomedical
application and to attract the practitioners of control
community. We divide the biomedical control
problem into three sections, rehabilitation
engineering, drug delivery and medical instrument. In
each section, we provide the reader with some
general background information and a brief
introduction of the recent results. The issues of each
application are also presented.
Keywords: Biocontrol, Biomedical Engineering

1. Introduction
The control mechanism composes basic for
maintenance of homeostasis at all levels of
organization in the hierarchy of living systems. The
control law of biologic system is evolved through
millions of years. It is perfect to balance every
physiologic state, and to maintain human life. From
the aging, the function of organs and tissues is
weakened little by little. The disease or accident can
also impair the function of living systems. Finally, the
control mechanism of physiology has broken and the
unstable biologic systems could endanger the life.
The goal of drug dosing and medical device assisting
is to recover natural control mechanism. It is hard to
completely replace natural control mechanism by
artificial method. So, to study the control theory in
biomedical engineering is necessary.
The objective of this article is to explore some
biomedical control problem and solution to general
control researcher. We wish to interest the control
community in the idea of developing applications in
* Corresponding author

medicine, and demonstrate to medical community

that control theory had solid applications in the
medical field. Besides, biomedical system provides
classic examples of effectively functioning control
system design. Medical topic studied regarding their
control structures and behavior can benefit in other
domains of control engineering practice.
Before going deeply into the biomedical control,
we should understand the difference between
engineering and physiological control systems. At
first, the engineering control system is designed for a
specific task, and we can fine-tune the controller
parameter to achieve optimal result. In contrast,
physiologic control systems are built for versatility,
and may be capable of several different functions.
Since the designer develops the engineering system,
the characteristics of the model are generally known.
On the other hand, the physiological system is
usually unknown and difficult to analyze. Moreover,
there exist serious cross-coupling among different
physiological control system. As for the engineering
systems, the state variables of model may be
independent (linear). The controller of physiological
systems, in general, is adaptive. This mean the
control system may be able to change the output not
only through feedback, but also by allowing the
controller characteristics change. And the engineering
system can only use the feedback controller to
achieve the objective. About the establishing of
feedback path, the feedback signal of engineering
control systems is explicitly subtracted from the
reference input, revealing clearly the using of
negative feedback. However, negative feedback in
the physiological system is embedded in the plant
characteristics. Table 1 gives a summary.
Table 1 Differences between engineering and
physiological control systems
Engineering control Physiological
systems
control systems
For a specific task
Built for versatility
Goal
Fine tuned parameter Capable of serving
Design
(optimal)
different functions
Generally known
Unknown and
Model
difficult to analyze
Independent
Cross-coupling
State
variables
Adaptive
Controller Feedback
Embedded
Feedback Explicitly

The organization of this paper can divide into

three sections, rehabilitation engineering, drug
delivery and medical instrument. In rehabilitation
engineering, we talk about the control of powered
prosthesis, functional electrical stimulation and
biofeedback control. In the section of drug delivery,
we discuss the model analysis and control strategy in
immune system, blood pressure regulation, blood
glucose control and anesthesia control. In the section
3, we also consider the control technique for
developing medical device

2. Rehabilitation engineering
Rehabilitation engineering can be described as
the device, method and application to meet the needs
of person with disabilities in engineering method.
Depend on the help of assist devices; disabilities can
improve the quality of living and return to original
life. In the following, we discuss some rehabilitation
technique, like powered prostheses, functional
electrical stimulation and biofeedback control.
2.1 The control of powered prostheses
The prostheses used in the limbs of the
individuals whom with amputations or congenitally
deficient extremities. In current technique, however,
prosthesis can be considered as a tool rather than a
limbs replacement. The real problem is the issue of
how to interface a multifunction arm or leg to an
amputee in a meaningful way. For this reason that the
prosthesis is often dominated by consideration of
control.
Electromyography (EMG) signal is the best
way to achieve multifunctional prosthesis control.
Englehart et al. have demonstrated the potential of
myoelectric base powered prosthesis control design
[1]. The fuzzy approach to classify single-site
electromyograph (EMG) signals for multifunctional
prosthesis control is also presented in [2]. This paper
focus on the classification problem, the ultimate goal
is to improve the control performance of myoelectric
prosthesis. The fuzzy approach was compared with
an artificial neural network (ANN) method on four
subjects, and a slightly superior classification results
were obtained.
We need to develop progress toward a more
natural, more effective means of myoelectric control
by providing high accuracy, low response time, and
an intuitive control interface to the user.
2.2 Functional electrical stimulation
There are quite difference between powered
prosthesis and functional electrical stimulation,
powered prosthesis use body signal to control
external assist device but function electrical
stimulation use external signal to control body.
Absence of basic functions such as grasping, walking,

breathing, and bladder voiding often renders such

subjects dependent on the assistance of others for
daily living activities. This cost the government lots
of money and efforts. Functional electrical
stimulation (FES) is the use of voltage pulses to
induce skeletal muscle contraction and consequently
joint movements. By precisely controlling electrical
stimulation, we can change the body postural and
accomplish many actions, which include balance,
standing and walking etc. FES has been successfully
adopted for stroke rehabilitation, spinal cord injuries
and many limbs motor dysfunction in clinic. Fig. 1
shows the block diagram of functional electrical
stimulation control.
EMG
Signal
Fource

FES output
Signal
Processing

Controller

channel 1
channel 2
channel N
controled
muscles

Fig. 1 Functional electrical stimulation control

block diagram
(1) Balance control
Balance is the most import, because stand, walk, and
many activities all rely on this ability. The frequency
of falls in elderly adults and Parkinsonian patients
implicates deterioration in the neural mechanisms
that govern postural stability, which is evident by
problems in maintaining both static and dynamic
balance. Wall et al. present body-tilt information to
the subject to prevent fall in the balance impaired[3].
The precursor prosthesis is a wearable, distributed,
modular design, which builds upon the single-axis
research device and consists of a 3-D motion sensor
array, a central processor, and vibrotactile stimulators.
Kuo also reported selection of control strategies used
by human in response to small perturbations to stable
upright balance[4]. A human postural optimal control
model is also provided to analyze and estimate
balance dynamic. There are several modes of
adaptation to postural perturbations used to withstand
balance disturbances and reduce the effects of muscle
fatigue[5].
(2) Stand and walk
Although most of us take it for granted,
walking is actually a complex task that requires
intricate neural control. Successful navigation
through our changing daily environments requires the
ability to adapt locomotor outputs to meet a variety of
situations. Jonic et al. use three supervised machine
learning (ML) techniques for prediction of the
activation patterns of muscles and sensory data, based
on the history of sensory data, for walking assisted by
FES[6]. Popovic et al. design an instrument to
improve quality of life in stroke/spinal cord injury
subjects with rapid prototyping and portable FES
systems[7]. One of the main features of both
stimulators is that they can reliably measure muscle
EMG activity and use this signal to trigger and
control the stimulation sequences. A patient-driven

control strategy for standing-up and sitting-down was

experimentally tested on two paraplegic patients by
applying functional electrical stimulation to the
quadriceps muscle[8] It is based on an inverse
dynamic model (IDM) that predicts the stimulation
pattern required to maintain the movement as it is
initiated by the patients voluntary effort.

delivery includes two model, pharmacokineticpharmacodynamics model and hemodynamic model.

Fig.
3
depicts
pharmacokinetics
and
pharmacodynamics as the fundamental elements of
pharmacology. In the following, we will discuss
some drug delivery applications, which involve
automatic technique to achieve the goal.
Drug
Dose

2.3 Biofeedback control

A major application for biofeedback is to provide
tools for detecting and controlling physiological
state, like heart rate, EEG and muscle activity. In
the control theory view, the controller of
biofeedback system is often patients brain. The
most import thing is to create a feedback path. Fig.
2 illustrates the three stages of biofeedback systems,
measurement, signal processing and presentation.

Drug
Effect

Pharmacokinetics

Drug
Concentration

Pharmacodynamics

Fig. 3 Bridge between pharmacokinetics and

pharmacodynamics
3.1 Immune system control: HIV example

audoi and visual

feedback

Feedback
presentation

Data
collection

Signal
conditioning

Human

Fig. 2 the biofeedback control system

In 1995, Moran et al. develop a biofeedback
cane system to measure axial cane force while
walking[9]. An audio alarm can be programming to
sound within the adjustable limits according to cane
load magnitude. Lee et al. provided a device to help
improve balance and postural control by using
biofeedback
for
standing-steadiness
and
weight-bearing training[10]. Wu also presented a
system that estimates and displays, in real-time, the
location of the center of gravity of the human body
relative to the feet was developed and then used in a
biofeedback training program for improving the
postural instability caused by deterioration of the
Proprioc eptive system in elderly patients with
significant diabetic sensory neuropathy during
perturbations of a support platform [11].
In biofeedback, the decision-making and
control computing is all accomplished by human
brain. So, The effect and performance of the
biofeedback is dominated in the accuracy of the
measured signal, the meaning of the signal processing
and the clear presentation.

3. Drug delivery for optimal therapy

Drug can be used to regulate physiological
variables, like blood pressure, blood glucose and
heart rate. Due to the toxicity and side effect of drug,
precise dosage regimens are important. In addition to,
timely drug delivery will help patient recovering
more quickly. Anesthesia is also an important drug
application in surgery.
The behavior of drug

Human immune system performs as an

ingenious control system. The control objective is to
resist external invader. Sometime immune system
may fail, we need to deliver drug recovering original
immune function. Over the last several years, most
significant progress has been made in our
understanding of human HIV infection. Maximal
suppression of the viral load to below detectable
levels has been achieved using highly active
antiretroviral therapy (HAART). HAART because
the high drug doses used have adverse side effects
that make adherence to therapy very difficult. A
regimen that could reduce dosage requirements while
maintaining control over viral plasma levels might
not only increase patient adherence but the overall
health of the patient by reducing side effects.
Jeffrey et al. showed an application of control
theory to human immunodeficiency virus (HIV)
models[12]. In this paper, minimum singular value
decomposition is applied to HIV models to measure
the extent to which the different stages in the
progression of HIV disease are controllable and,
consequently, when best to initiate therapy such that
the general objectives of therapy are satisfied[12].
Brandt et al. describes a continuous differential
equation model of the interaction dynamics of HIV-1
and CD4 and CD8 lymphocytes in the human body.
They also demonstrate several methods of stable
control of the HIV-1 population using an external
feedback control term that is analogous to the
introduction of a therapeutic drug regimen.
None of these immune models, however, can
completely exhibit all that is observed clinically and
account for the full course of the disease. The main
reason for the models limitation is lack of a good
understanding of the immunology of the human body
against
HIV.
Biological
systems
exhibit
multicompartmental interactions that are usually not
well understood and as a result, cannot be accurately
modeled mathematically.

3.2 Blood pressure regulation

In order to reduce the blood lost during
operation, a blood pressure regulation system is
necessary. The main method is the infusion of
sodium nitroprusside in order to lower blood pressure
in patients who have undergone surgery. In tradition,
the bolus injection can rapidly decrease blood
pressure, but has disadvantage that the effect
diminishes rapidly and it can only be applied
periodically in order to avoid cyanide poisoning.
Therefore, the idea of continuously controlled release
of the drug has been proposed. The controlled release
method has the advantage of achieving lower blood
pressures over longer periods of time. The control
problem is to find the correct dose, which quickly
lowers the blood pressure to the desired level, while
avoiding a drug overdose.
Adapting several parameters, especially in a
nonlinear controller, can lead to undesired behavior
because it is non-trivial to predict the closed loop
system response for the whole operating range and
for any possible parameter variation. In the paper[13],
analysis of the model parameters has shown that an
internal model controller (IMC) that can meet robust
stability and performance criteria can be designed for
variations in all of the parameters. Furutani et al.
developed a state predictive controller to cope with
the dead time existing in the responses for the drug
delivery[14]. The adaptive control is also used in
mean arterial pressure through the intravenous
infusion of sodium nitroprusside[15;16].
Many different controllers have been designed
for the problem of blood pressure regulation and most
of them adapt several parameters in order to deal with
the uncertainties of the system. However, due to the
fact that the ultimate goal is to design a controller that
can be used in a clinical environment, the controller
should be as simple as possible.
3.3 Blood glucose regulation
Diabetes mellitus is a disease characterized by
the inability of the pancreas to regulate blood glucose
concentration., such that exogenous insulin is
required to control the disease. Almost all
insulin-dependent diabetic subjects live with a
conventional or intensified insulin therapy regimen.
In order to improve the efficiency of dosing, the aim
is to establish closed-loop control of blood sugar
level (BSL), mimicking the endocrine pancreas. In
general, a closed-loop feedback system for insulin
delivery consists of a blood glucose sensor, a
controller, and an infusion pump[17;18]. The fig 4
and fig 5 display the control diagram in partially
closed-loop control strategy and real-time closed loop
control strategy. It is revealed that partially control
needs manual intervention in glucose measurement
and insulin injection. But real-time control with
subcutaneous route insulin injection is fully
automatic. This method is feasible in patients in

intensive care.
The paper published by Chees et al. show the
difference between manual control and closed-loop
system control, real-time to control blood sugar level
in patients with critical illness[19]. A closed-loop
control system was constructed to use continuous
glucose monitoring system in a real-time manner,
coupled with a proportional integral control algorithm
based on a sliding scale approach, for automatic
intravenous infusion of insulin to patients. Neural
prediction control had been used in closed loop
control using subcutaneous glucose measurement and
injection[20]. The result also demonstrates that the
real-time control of blood glucose is feasible.
Doctor's
advics
+

Input
command Feedback
control

SC
injection

Patient

Blood glucose
measurement

Fig 4 partially closed-loop control strategy of

conventional insulin therapy
Set
Point

Controller

SC Injecton
Pump

Patient

SC glucose
measurement

Fig 5 closed-loop strategy with subcutaneous route

insulin injection
The primary need for constructing an artificial
pancreas is as a reliable, long-term in vivo sensor for
glucose concentration, a device that is currently
unavailable. A key tenet from robust control theory is
that controller performance is directly linked to
model accuracy. The issues such as time delay and
sensor dynamics directly affect closed-loop
performance. Ongoing work is exploring methods for
capturing variations in the nonlinear patient model
using the linear model with variable parameters and
measurement data. Fault detection is another
algorithmic issue to be addressed.
3.4 Anesthesia control
Adequate anesthesia can be defined as a
reversible pharmacological state in which the
patients muscle relaxation, analgesia, and hypnosis
are guaranteed. Fig 6 shows the input/output relation
of the anesthesia problem. From this figure, we can
observe that anesthesia control is a complicate
problem. Thus, anesthesiologists adopt the role of a
feedback controller. The uses of automatic anesthesia
controllers are capable of taking over and improve
parts of such a complex decision process. If the
routine tasks are taken over by automatic controllers,
anesthesiologists are able to concentrate on critical
issues that may threaten the patients safety, the
automatic controllers would be able to provide drug

administration and avoid overdosing. The ultimate

advantage would be a reduction in costs due to the
reduced drug consumption and the shorter time spent
by the patient in the postanesthesia care unit.
I.V. Anesthetics

Hypnosis

Volatile Anesthrtics

Analgesia

Muscle Relaxants

Relaxation

Ventilation Parameters
Nacl
Surgical stimulus
Blood loss

Human
Body

EEG
Heart rate
CO2
Bloos pressure

Fig. 6 The input/output relation of the anesthesia

problem
Automatic controllers capable of regulating
multiple patient outputs for higher-quality anesthesia
treatment are discussed in [21]. Maplenson et al.
explore the identification problem when dealing with
physiological models relating to anaesthetic
drugs[22]. The Mapleson model for drug
concentration is described by algebraic equations,
which are derived from the laws of physics and
chemistry. Gentilini had presented a model-based
closed-loop control system to regulate hypnosis with
the volatile anesthetic isoflurane [23]. Hypnosis is
assessed by means of the bispectral index (BIS), a
processed
parameter
derived
from
the
electroencephalogram. There are also many articles
assessed hypnosis by using auditory evoked potential
[24;25]
The controllers cannot be used without the
anesthesiologists close supervision. The feedback
systems that were proposed are SISO, whereas the
goal of the anesthesiologist is to maintain several
physiological variables in specified ranges. Further,
the benefits of multidrug anesthesia are not yet
exploited. At present, Isoflurane is used alone, and
other drugs in the control schemes are not considered.

4. Medical instrumentation technique

In this section, we will focus the work on using
external device to present exogenous feedback
regulation.
4.1 Temperature and humidity regulation
The temperature and the humidity are an
important environment parameters, which not only
affect our feeling, but also change our physiology.
The temperature in the air could affect the lost rate of
heat from skin. For this reason, many physiological
parameters like body temperature, heart rate and
humidity of skin vary by temperature and humidity.
In medical intensive care condition, the requirement
of environment parameter regulation is very crucial.
The variation of temperature and humidity is

conditioned by each other.

Radiant warmers and incubators are used to
maintain the body temperature of newborn infants.
But radiant warmers increase convective and
evaporative heat loss and insensible water loss.
Bouattoura et al. proposed an active humidification
system to achieve high and steady humidity
levels[26]. The algorithm is based on a combination
of optimal control theory and dynamic programming
approach. Temperature control can also influence
some physiological variables, like heart rate, breath
and anesthesia levels. Qiu et al. developed
temperature control system for magnetic resonance
microscopy[27]. With control of body temperature,
heart rate is stabilized and repetition time during
cardiac-gated studies is less variable. Thus, image
quality and resolution are improved.
Accurate control is made not easily. The reason
is that the level of humidity achieved is seldom
related linearly to the control setting. Regulation is
made even more difficult by the absence of any
device for measuring humidity. The automated
control system allows us more quickly to warm the
subject to a target temperature while avoiding
significant temperature overshoot with minimal
subject deviations about the set point.
4.2 Surgery assist device
In this section, we will introduce two surgery
assist devices, electrosurgery scalpel and robot assist.
Electrosurgery is a minimally invasive medical
technique compared with conventional surgery. It
consists of three functions, to destroy benign and
malignant lesions, control bleeding and cut tissue.
The primary cutting effect is accompanied by
coagulation and hence hemostasis. The scalpel and
the tissues are in ohmic contact. Then the electrical
device generates radio frequency current from
electrode to tissues. High heat is produced by large
electrical power dissipating in ohmic contact. We
have to control the power passing to tissues carefully.
The task is now to determine continuously the
appropriate electrical power, since this depends not
only on the kind of tissue but also on the type of
electrode used. Robotic assistants aid surgeons is also
a minimally invasive technique. A robotic system for
collaboratively
or
autonomously
performing
endoscopic procedures is the most common method.
By eliminating the large incision and extensive
dissection, much of the pain of recovery can also be
eliminated and the length of hospital stay reduced.
Kang et al. presented the design and
implementation of a new robotic system for assisting
surgeons in performing minimally invasive surgical
procedures[28]. This system is designed for
collaborative operation between the surgeon and the
robot. A new approach in radio frequency
electrosurgery, used for tissue treatment, is achieved
by using a new process control method[29]. An

external control unit allows a commonly available

rf-generator to automatically supply the appropriate
power for differing tissue types, thus ensuring best
cutting quality. The sparks, generated during the
scalpel electrode interaction with the tissue, appear
statistically distributed.
In terms of further technical development of
surgical robot, one focuses on making the
autonomous surgical procedures more robust. Careful
modeling and control of the thread tension is critical
in avoiding tissue tearing and pulling the needle from
the stitcher or the holder.
4.3 Artificial hearts
Cardiac assist device is used to replace the
natural heart and to wait for available organ donor. It
can be lowered the load of impaired heart. We need
to control the artificial heart provide enough flow rate
to meet the requirement of patients activity.
An automatic physiologic control system for
the actively filled, alternately pumped ventricles, in
long-term use had been developed by Kim et al. [30].
The automatic control system ensure the device
maintains a physiologic respond for cardiac output,
compensate any nonphysiological condition and is
stable, reliable and operate in high power efficiency.
Maslen et al. given a brief survey of the artificial
hearts reveals that its underlying premise is to
develop active augmentations to the human biological
function[31]. The system identification method is
used for developing heart model for the control of a
cardiac ventricular assist device[32;33].
We can find two control issues in developing
an adequate artificial heart. (1) The power
consumption of artificial heart must as low as
possible, because the heat result from the dissipation
of pump and electron could depress biocompatibility.
Besides, the good efficiency can permit portable
requirement
from
battery
with
suitable
charge-to-charge cycle. (2) Design a flow control
system that ensures that the patient blood pressure is
appropriate to his level of activity.

5. Conclusion
As this work shows, the requirements of
control theory in biomedical application are
extremely indispensable. The involving of control
theory makes biomedical application more efficient
and more natural. This paper clearly indicates that
considerable progress is being achieved in addressing
these challenging problem of biomedical control
which are relevant to the needs of the community of
control practitioners. There are of course many areas
that are not represented. One of the main sets of the
work that is missing is the behavior of neuron and the
brain. Neural modulation is a very modern idea to
treat the epilepsy and Parkinsons disease. The
technique
in
engineering
can
provide

neuronmodulation a proper signal processing method.

We hope to call more peoples attention of control
community in the idea of biomedical applications. It
is believed that as more and more control researches
are proposed, the ultimate solutions in biomedical
applications are available.

References
[1] Englehart, K. and Hudgins, B., "A robust,
real-time control scheme for multifunction
myoelectric control," IEEE Trans.Biomed.Eng.,
vol. 50, no. 7, pp. 848-854, 2003.
[2] Chan, F. H. Y., Yong-Sheng, Y., Lam, F. K.,
Yuan-Ting, Z., and Parker, P. A., "Fuzzy EMG
classification for prosthesis control," IEEE
Trans.Rehab.Eng., vol. 8, no. 3, pp. 305-311,
2000.
[3] Wall, C., III and Weinberg, M. S., "Balance
prostheses for postural control," IEEE Eng.in
Med.and Biol.Mag., vol. 22, no. 2, pp. 84-90,
2003.
[4] Kuo, A. D., "An optimal control model for
analyzing human postural balance," IEEE
Trans.Biomed.Eng., vol. 42, no. 1, pp. 87-101,
1995.
[5] Fransson, P. A., Johansson, R., Tjernstrom, F.,
and Magnusson, M., "Adaptation to vibratory
perturbations in postural control," IEEE Eng.in
Med.and Biol.Mag., vol. 22, no. 2, pp. 53-57,
2003.
[6] Jonic, S., Jankovic, T., Gajic, V., and Popvic, D.,
"Three machine learning techniques for
automatic determination of rules to control
locomotion," IEEE Trans.Biomed.Eng., vol. 46,
no. 3, pp. 300-310, 1999.
[7] Popovic, M. R., Keller, T., Papas, I. P. I., Dietz,
V., and Morari, M., "Surface-stimulation
technology for grasping and walking
neuroprostheses," IEEE Eng.in Med.and
Biol.Mag., vol. 20, no. 1, pp. 82-93, 2001.
[8] Riener, R., Ferrarin, M., Pavan, E. E., and Frigo,
C. A., "Patient-driven control of FES-supported
standing up and sitting down: experimental
results," IEEE Trans.Rehab.Eng., vol. 8, no. 4,
pp. 523-529, 2000.
[9] Moran, P. T., Harris, G. F., Acharya, K.,
Hongsheng, Z., and Wertsch, J. J., "A
biofeedback cane system: instrumentation and
subject application results," IEEE
Trans.Rehab.Eng., vol. 3, no. 1, pp. 132-138,
1995.
[10] Lee, M. Y., Wong, M. K., and Tang, F. T.,
"Using biofeedback for standing-steadiness,
weight-bearing training," IEEE Eng.in Med.and
Biol.Mag., vol. 15, no. 6, pp. 112-116, 1996.
[11] Ge, W., "Real-time feedback of body center of
gravity for postural training of elderly patients
with peripheral neuropathy," IEEE
Trans.Rehab.Eng., vol. 5, no. 4, pp. 399-402,

1997.
[12] Jeffrey, A. M., Xiaohua, X., and Craig, I. K.,
"When to initiate hiv therapy: a control theoretic
approach," IEEE Trans.Biomed.Eng., vol. 50, no.
11, pp. 1213-1220, 2003.
[13] Hahn, J., Edison, T., and Edgar, T. F., "Adaptive
IMC control for drug infusion for biological
systems," Control Eng.Practice, vol. 10, no. 1,
pp. 45-56, 2002.
[14] Furutani, E., Araki, M., and Maetani, S., "Blood
pressure control during surgical operations,"
IEEE Trans.Biomed.Eng., vol. 42, no. 10, pp.
999-1006, 1995.
[15] Maitelli, A. L. and Yoneyama, T., "Suboptimal
dual adaptive control for blood pressure
management," IEEE Trans.Biomed.Eng., vol. 44,
no. 6, pp. 486-492, 1997.
[16] Polycarpou, M. M. and Conway, J. Y., "Indirect
adaptive nonlinear control of drug delivery
systems," IEEE Trans.Automat.Contr., vol. 43,
no. 6, pp. 849-856, 1998.
[17] Parker, R. S., Doyle, F. J., III, and Peppas, N. A.,
"The intravenous route to blood glucose control,"
IEEE Eng.in Med.and Biol.Mag., vol. 20, no. 1,
pp. 65-73, 2001.
[18] Bellazzi, R., Nucci, G., and Cobelli, C., "The
subcutaneous route to insulin dependent diabetes
therapy," IEEE Eng.in Med.and Biol.Mag., vol.
20, no. 1, pp. 54-64, 2001.
[19] Chee, F., Fernando, T., and van Heerden, P. V.,
"Closed-loop glucose control in critically ill
patients using continuous glucose monitoring
system (CGMS) in real time," Information
Technology in Biomedicine, IEEE Transactions
on, vol. 7, no. 1, pp. 43-53, 2003.
[20] Trajanoski, Z., Regittnig, W., and Wach, P.,
"Neural predictive controller for closed-loop
control of glucose using the subcutaneous route:
a simulation study," Control Eng.Practice, vol. 5,
no. 12, pp. 1727-1730, 1997.
[21] Gentilini, A., Frei, C. W., Glattfedler, A. H.,
Morari, M., Sieber, T. J., Wymann, R., Schnider,
T. W., and Zbinden, A. M., "Multitasked
closed-loop control in anesthesia," IEEE Eng.in
Med.and Biol.Mag., vol. 20, no. 1, pp. 39-53,
2001.
[22] Mahfouf, M., Linkens, D. A., and Xue, D., "A
new generic approach to model reduction for
complex physiologically based drug models,"
Control Eng.Practice, vol. 10, no. 1, pp. 67-81,
2002.
[23] Gentilini, A., Rossoni-Gerosa, M., Frei, C. W.,
Wymann, R., Morari, M., Zbinden, A. M., and
Schnider, T. W., "Modeling and closed-loop
control of hypnosis by means of bispectral index
(BIS) with isoflurane," IEEE Trans.Biomed.Eng.,
vol. 48, no. 8, pp. 874-889, 2001.
[24] Linkens, D. A., Abbod, M. F., and Backory, J.
K., "Closed-loop control of depth of
anaesthesia:a simulation study using auditory

evoked responses," Control Eng.Practice, vol. 5,

no. 12, pp. 1717-1726, 1997.
[25] Huang, J. W., Ying-Ying, L., Nayak, A., and
Roy, R. J., "Depth of anesthesia estimation and
control [using auditory evoked potentials],"
IEEE Trans.Biomed.Eng., vol. 46, no. 1, pp.
71-81, 1999.
[26] Bouattoura, D., Villon, P., and Farges, G.,
"Dynamic programming approach for newborn's
incubator humidity control," IEEE
Trans.Biomed.Eng., vol. 45, no. 1, pp. 48-55,
1998.
[27] Qiu, H. H., Cofer, G. P., Hedlund, L. W., and
Johnson, G. A., "Automated feedback control of
body temperature for small animal studies with
MR microscopy," IEEE Trans.Biomed.Eng., vol.
44, no. 11, pp. 1107-1113, 1997.
[28] Hyosig, K. and Wen, J. T., "Robotic assistants
aid surgeons during minimally invasive
procedures," IEEE Eng.in Med.and Biol.Mag.,
vol. 20, no. 1, pp. 94-104, 2001.
[29] Wurzer, H., Maeckel, R., Lademann, J.,
Audring, H., and Liess, H. D., "A spark counter
as a control unit of a radio frequency surgery
device," IEEE Trans.Biomed.Eng., vol. 44, no. 9,
pp. 831-838, 1997.
[30] Hee, C. K., Khanwilkar, P. S., Bearnson, G. B.,
and Olsen, D. B., "Development of a
microcontroller-based automatic control system
for the electrohydraulic total artificial heart,"
IEEE Trans.Biomed.Eng., vol. 44, no. 1, pp.
77-89, 1997.
[31] Maslen, E. H., Bearnson, G. B., Allaire, P. E.,
Flack, R. D., Baloh, M., Hilton, E., Noh, M. D.,
Olsen, D. B., Khanwilkar, P. S., and Long, J. D.,
"Feedback control applications in artificial
hearts," Control Systems Magazine, IEEE, vol.
18, no. 6, pp. 26-34, 1998.
[32] Yih-Choung, Y., Boston, J. R., Simaan, M. A.,
and Antaki, J. F., "Estimation of systemic
vascular bed parameters for artificial heart
control," IEEE Trans.Automat.Contr., vol. 43, no.
6, pp. 765-778, 1998.
[33] Johansson, R., Holm, M., Olsson, S. B., and
Brandt, J., "System identification of atrial
activation during chronic atrial fibrillation in
man," IEEE Trans.Automat.Contr., vol. 43, no. 6,
pp. 790-799, 1998.